Re: Bug with pg_basebackup and 'shared' tablespace

On Fri, Sep 29, 2017 at 2:19 PM, Kyotaro HORIGUCHI
<horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> wrote:
> It would practically work but I don't like the fact that the
> patch relies on the specific directory/file ordering in the tar
> stream. This is not about the CATVER directory but lower
> directories. Being more strict, it actually makes excessive calls
> to verify_dir_is_em..() for more lower directories contrarily to
> expectations.
>
> I think that we can take more more robust way to detect the
> CATVER directory. Just checking if it is a top-level directory
> will work. That is, making the following change.

My tendency about this patch is still that it should be rejected. This
is presenting additional handling for no real gain.

>
> - if (firstfile && !basetablespace)
> + /* copybuf must contain at least one '/' here */
> + if (!basetablespace && strchr(copybuf, '/')[1] == 0)
>
> This condition exactly hits only CATVER directories not being
> disturbed by file ordering of the tar stream.

Anyway, as a new version is at least needed, I am marking it as
returned with feedback. The CF is close to its end.
--
Michael